Multi step forecast of multiple time series at once in Python (or R)

Question

I have problem quite similar to M5 Competition - i.e. hierarchical data of many related items.

I am looking for best solution where I can forecast N related time series in one run. I would love to allow model to learn internal dependencies between each time series in the run.

I am aware I can use Darts or TeporalFusionTransfomer (with pythorch) however I am wondering whether there is some more .... universal approach for example where I can choose N models or I can implement external models ( like XGBoost / LGBM )

Also I am wondering how to align it with using Quantile loss function.

I am somehow new to the problem of hierarchical /// simultaneous forecasting problem, so each piece of advice // materials to learn will be extremely helpful.